Predictingfive Currencyexhchange Rates Using Five Different Learning Algorithms of Neural Network: a Case Study

One of the most important hopes in modern finance is finding the most accurate ways to forecast future values of exchange rates.The research provides some evidence about the efficiency of utilizing neural network models in forecasting foreign exchange rates. We used three distinct learning algorithms in our neural network models, namely,Scaled Conjugate Gradient (SCG), Standard Backpropagation (SBP), Bckpropagation with Bayesian Regularization (BPR), Gradient Descent with Momentum (GDM) and resilientBackpropagation (RB); which were used for Unites State Foreign Exchange to predict five diverse currencies against United State dollar. We chose four most important technical indicators based on the Forex experts' opinion to feed neural network models. Finally, a general comparison between neural network models and a BoxJenkins model (ARIMA) forecasting model based on the five performance metrics is offered.